LHC@home

Hi all,

We are testing out a new app which allows you to run much longer ATLAS simulation tasks. These tasks run over 1000 events instead of the usual 200 and so will take 5 times longer. The input data download is the same but the output data to upload is 5 times larger.

These tasks are designed for large powerful machines which are dedicated to running ATLAS and the aim is to reduce the load on the BOINC servers from large clusters currently running ATLAS tasks, and allow larger clusters to join in the future. They are not suitable for background running or less powerful hosts so please think carefully before trying them out! We will of course keep the regular shorter tasks going.

The app is currently only available for native Linux users, and requires minimum 4 cores per task, however you may run up to 48-cores per task.

To try it out, select "ATLAS (long simulation)" in your LHC@Home preferences, and since it is still in the testing phase you must select "Run test applications" too.

Note that the first batch of tasks will be for test purposes and the data produced will not be used, so if you want your machines to produce only data useful for science then wait until the real batches start :)

Let us know of any issues or feedback and happy crunching!

https://lhcathome.cern.ch/lhcathome/workunit.php?wuid=158521681
CentOS8-VM with squid 6 CPU's

There was a minor configuration issue with 1.00 so we just released 1.01.

I think there's still a configuration problem...

This is the host https://lhcathome.cern.ch/lhcathome/show_host_detail.php?hostid=10686009

It is an Intel i5-4670K, 4c/4t Haswell CPU, 24 GM memory, running Linux Debian 10 (Buster). It's native Linux, not a VM.

I was using it yesterday to successfully run the long ATLAS tasks on the development server.

Relevant snippet from the client log:

3/25/2021 10:25:23 AM | LHC@home | Sending scheduler request: To fetch work.
3/25/2021 10:25:23 AM | LHC@home | Requesting new tasks for CPU
3/25/2021 10:25:25 AM | LHC@home | Scheduler request completed: got 0 new tasks
3/25/2021 10:25:25 AM | LHC@home | No tasks sent
3/25/2021 10:25:25 AM | LHC@home | No tasks are available for ATLAS (long simulation)
3/25/2021 10:25:25 AM | LHC@home | Message from server: VirtualBox is not installed

(The vbox warning probably isn't important.)

There's two things I know are NOT the problem.

1) It's not my preferences, because another computer assigned to the same venue (default) just got a long ATLAS task.
2) I didn't break this computer's configuration after testing on the dev system, because I moved it to running the production short ATLAS tests and it got a task successfully.

The only obvious conclusion is that there's something wrong in the server configuration, most likely the plan class. The two machines have different CPUs (Intel vs. AMD), different cores (4 vs 8), and different memory (24 GB vs. 32 GB.)

EDIT: Another difference is that the computer that is able to get tasks is a Linux VM running on a Windows host, while the computer that is not able to get tasks is not a VM. Neither system has VBOX installed.

---

Want to find one of the largest known primes? Try PrimeGrid. Or help cure disease at WCG.

There are/were just a few long tasks in the queue.
Might be all of them are already sent out.

Or:
You forgot to activate the "beta app" box in your preferences:
https://lhcathome.cern.ch/lhcathome/prefs.php?subset=project

There are/were just a few long tasks in the queue.
Might be all of them are already sent out.

Or:
You forgot to activate the "beta app" box in your preferences:
https://lhcathome.cern.ch/lhcathome/prefs.php?subset=project

All is GOOD on your end.

There was an ID 10T error. :)

I realized something wasn't right when the short Atlas task was running only on one core. I had set that machine to run only on 1 core while trying to get Theory (native) to run on it (before understanding why it didn't, and giving up.) The long tasks won't download with less than 4 cores, so problem solved. My bad!

---

Want to find one of the largest known primes? Try PrimeGrid. Or help cure disease at WCG.

Worked just fine with 48 threads.

https://lhcathome.cern.ch/lhcathome/result.php?resultid=307102515

Since midnight no more longrunner, but Server-Staus show Tasks avalaible.
https://lhcathome.cern.ch/lhcathome/server_status.php

Cmpleted two Atlas tasks on my Linux laptop with good results. Got two more.
Tullio

---

Cmpleted two Atlas tasks on my Linux laptop with good results. Got two more.
Tullio

Tullio,
Your mentioned tasks are not the long ATLAS simulations meant in this thread and even not 'native', but the 'normal' VBOX ATLAS tasks.

Long Simulation is back.

Yes I know. But I have only two cores and cannot get them.In QuChemPedIA@home I am competing with hosts having more than 124 processors and my Intel I5 9400F has only three cores and 6 processors yet my RAC ranking, using VirtualBox on a Linux native project is 27.
Tullio

---

Tullio,
your both Win10 PC's have 6 or 8 CPU's for the longrunner from Atlas, when you build a Linux-VM.

I have a Linux VM on a Windows 10 PC running SuSE Thumbleweed with kernel 5.11..6-1-default But it is updated so frequently that I have to reboot it almost daily.
Tulliu

---

Well am still trying to get one of these long ones lol

Sooner or later, I will get one.

Cheers

Did you activate "Run test applications?" for the venue your computer is attached to?
https://lhcathome.cern.ch/lhcathome/prefs.php?subset=project

Even then you may still need a couple of attempts.

Hi computezrmle,

Well that could be it, will have a look when i get home from work.

Cheers

Hi computezrmle

That did it, came home and found one crunching away.

Thankyou

Are you aware that your CVMFS is misconfigured?

[2021-03-29 08:58:28] VERSION PID UPTIME(M) MEM(K) REVISION EXPIRES(M) NOCATALOGS CACHEUSE(K) CACHEMAX(K) NOFDUSE NOFDMAX NOIOERR NOOPEN HITRATE(%) RX(K) SPEED(K/S) HOST PROXY ONLINE
[2021-03-29 08:58:28] 2.9.0.0 2551 808 62280 81656 0 62 3785025 4194305 2 130560 0 369983 99.977 17734 5 http://s1asgc-cvmfs.openhtc.io:8080/cvmfs/atlas.cern.ch http://128.142.248.156:3126 1

This points out that all requests are send to a CERN fall-back proxy in Geneva (128.142.248.156).
From there they go to openhtc.io in Europe.


Since your cluster reports much more than 5 worker nodes (=cores) the recommended way would be to use a local HTTP proxy.
Second best would be to use openhtc.io DIRECT and switch off the fallback-proxies.
This would ensure to get the data from openhtc.io proxies in Australia rather than Europe.